Movatterモバイル変換


[0]ホーム

URL:


CN1536842A - Equipment for controlling access of facilities according to the type of application - Google Patents

Equipment for controlling access of facilities according to the type of application
Download PDF

Info

Publication number
CN1536842A
CN1536842ACNA2004100325476ACN200410032547ACN1536842ACN 1536842 ACN1536842 ACN 1536842ACN A2004100325476 ACNA2004100325476 ACN A2004100325476ACN 200410032547 ACN200410032547 ACN 200410032547ACN 1536842 ACN1536842 ACN 1536842A
Authority
CN
China
Prior art keywords
memory
access
class
address
queue
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CNA2004100325476A
Other languages
Chinese (zh)
Other versions
CN1310475C (en
Inventor
理查德・L・阿恩特
理查德·L·阿恩特
・L・比克马
布鲁斯·L·比克马
F・克拉多克
戴维·F·克拉多克
・E・富斯
罗纳德·E·富斯
・A・格雷格
托马斯·A·格雷格
・M・沃克
布鲁斯·M·沃克
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
International Business Machines Corp
Original Assignee
International Business Machines Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by International Business Machines CorpfiledCriticalInternational Business Machines Corp
Publication of CN1536842ApublicationCriticalpatent/CN1536842A/en
Application grantedgrantedCritical
Publication of CN1310475CpublicationCriticalpatent/CN1310475C/en
Anticipated expirationlegal-statusCritical
Expired - Fee Relatedlegal-statusCriticalCurrent

Links

Images

Classifications

Landscapes

Abstract

An apparatus, system and method for controlling access to facilities based on usage class of a requestor are provided. With the apparatus, system and method, a two level protection mechanism is provided for protecting host channel adapter (HCA) facilities from unauthorized access. With the present invention, a first level of access is provided through virtual address translation and a mechanism for determining if the requestor of access may access a system memory address space page associated with a real address to which the virtual address maps. A second level of access is provided through the allocation of usage classes and determining a required usage class for accessing an HCA facility.

Description

According to classes for access to the facility to control a device, system and method
Technology
The invention relates to a class according to (usage class) to control access to the loading facility Set, systems and methods. More specifically, the present invention provides a method for the access requester according to the phase Associated with the use of classes to control logic portion associated with the host channel adapter facilities accessible machine The system.
BACKGROUND
Such as InfiniBand (IB) network system area network (SAN), the hardware provides the consumer Information transfer mechanism, which can be used for input / output devices (I / O), and the process between general computing nodes Interprocess communication (IPC). Processes executing on the device to the SAN through the channel adapter (CA) on Send / receive work queue register (posting) to send / receive messages to messaging access the SAN hardware. These processes are also called "consumers."
Such as InfiniBand (IB) network system area network (SAN), the hardware provides the consumer Information transfer mechanism, which can be used for input / output devices (I / O), and the process between general computing nodes Interprocess communication (IPC). Processes executing on the device to the SAN through the channel adapter (CA) on Send / receive work queue register (posting) to send / receive messages to messaging access the SAN hardware. These processes are also called "consumers."...
Such as InfiniBand (IB) network system area network (SAN), the hardware provides the consumer Information transfer mechanism, which can be used for input / output devices (I / O), and the process between general computing nodes Interprocess communication (IPC). Processes executing on the device to the SAN through the channel adapter (CA) on Send / receive work queue register (posting) to send / receive messages to messaging access the SAN hardware. These processes are also called "consumers."...
Target channel adapter (TCA) is used as a channel adapter from the main topic of the message sent by the node Point of use. Target channel adapter and the main channel adapter to the target node provides access to configuration of the SAN Q points have similar functions.
Target channel adapter (TCA) is used as a channel adapter from the main topic of the message sent by the node Point of use. Target channel adapter and the main channel adapter to the target node provides access to configuration of the SAN Q points have similar functions....
In general, a logical partition system to support multiple operating system environment is beneficial. Such logical partition allows each operating system and operating in the operating system environment applications You can verify that the operating system on a dedicated system resources, even if the resource is actually multiple operating Operating system environment sharing. Such a system allows logical partitions can support multiple different environments, Therefore, by sharing resources in different environments reduce maintenance costs and improve the system of common Sex.
Although the logical partitions are well known in the art, but the logical partitions have not been applied to the system Area network, such as the InifiniBand network structure. Moreover, in the present invention do not know how to achieve before Have been logical partition in areas of the system resources in the network access control for the operating system and applications A program can not access other logical partitions are assigned to resources. Therefore, advantageously, is a one kinds Devices, systems and methods to achieve such InifiniBand area network system, the network logic of Area, for the network associated with the logical partition is controlled access to resources.
SUMMARY OF THE INVENTION
The present invention provides for the use of classes according to the requestor access to the facility to control a device, system Systems and methods. The apparatus of the present invention, systems and methods, a two-stage type is provided with a protection mechanism The protection of the main channel adapter (HCA) facility to prevent unauthorized access. The present invention, Virtual address translation and a mechanism to provide access to the first level, the mechanism used to determine the access request Requester can access the mapped virtual address associated with the actual address of the system memory address space Page. Providing a second level access by using the class distribution and determining the necessary facilities for accessing the HCA Use classes. ...
The present invention provides for the use of classes according to the requestor access to the facility to control a device, system Systems and methods. The apparatus of the present invention, systems and methods, a two-stage type is provided with a protection mechanism The protection of the main channel adapter (HCA) facility to prevent unauthorized access. The present invention, Virtual address translation and a mechanism to provide access to the first level, the mechanism used to determine the access request Requester can access the mapped virtual address associated with the actual address of the system memory address space Page. Providing a second level access by using the class distribution and determining the necessary facilities for accessing the HCA Use classes. ...
In the appended claims that a given feature of the invention The novel features. However, through the Over the following detailed description with accompanying drawings the illustrated embodiment, the present invention itself, as well as preferred mode of use, and Additional objects and advantages which will be best understood.
Figure 1 is a diagram according to the present invention, a preferred embodiment of a distributed computer system diagram;
Figure 2 is according to the present invention, a preferred embodiment of the functional block diagram of the host processor node;
Figure 3A according to the present invention, a preferred embodiment of the adapter of FIG main channel;
3B is according to the present invention, a preferred embodiment of the switch of FIG.;
Figure 3C according to the present invention, a preferred embodiment of FIG router;
Figure 4 is a diagram according to the present invention, a preferred embodiment of the process in Fig work request;
Figure 5 is a diagram using a reliable connection service according to the present invention, a preferred embodiment of a distributed A part of a computer system diagram;
Figure 6 is a diagram in accordance with the use of a reliable datagram service connection with a preferred embodiment of the invention Example The portion of the distributed computer system; Fig
Figure 7 according to the present invention, a preferred embodiment of a data packet; Fig.
Figure 8 according to the present invention, a preferred embodiment of a portion of distributed computer system Figure;
Figure 9 is a diagram according to the present invention is used in a distributed networked system of FIG network addressing;
Figure 10 is a configuration diagram SAN subnet in the illustration of the invention the structure of a preferred embodiment of the , The portion of the distributed computing system diagram;
Figure 11 is a preferred embodiment of the present invention used in the example structure of the hierarchical communication;
Figure 12 is a diagram according to the present invention, logical partitions schematic view of the main channel adapter;
Figure 13 is a memory space according to the present invention, a schematic view;
Figure 14 illustrates in accordance with the present invention, the queue assigned to the demonstration of the system memory address space page Expanded view of the surface;
Figure 15 illustrates when determining whether there should be access to the HCA facilities provided by the requester the HCA Access to the facility, the invention is an exemplary embodiment of the overall operation;
Figure 16 is in accordance with the present invention a preferred embodiment, is used to determine whether it should be granted the HCA Access to the facility exemplary flowchart of a process.
Specific embodiments
The present invention provides a method for the control according to the class of the device access to system resources, the Department of Systems and methods. The preferred embodiment of the present invention will be described by InifiniBand architecture, InifiniBand Architecture of the present invention can be used as a possible system area network. Therefore, with reference to Figure 1-11 to provide For instructions on InfiniBand network.
Referring now to the drawings, in particular to Figure 1, which illustrates the present invention according to a preferred The distributed computer system of Example FIG. In Figure 1 represents a computer system with a distributed system Area Network (SAN) 100 form, and is provided to merely for illustrative purposes, the following Embodiments of the invention may be in a variety of other types and configurations to be implemented on a computer system. For example, the real The computer system of the present invention can range from having a processor and a few input / output (I / O) The adapter to a small server with thousands of processors and thousands of I / O adapter as a whole Parallel super computer system. Moreover, the present invention may be implemented by the Internet or an intranet connection far Remote computer system infrastructure. ...
SAN 100 is a high-bandwidth, low-latency event network, which interconnects the distributed computer system Node. A node is attached to one or more of the network links in the network and a message is formed The starting and / or destination of any parts. In the example described, SAN 100 includes the following form Node: node main processor 102, a main processor node 104, redundant array of independent disk (RAID) Subsystem node 106 and I / O backplane node 108. In Figure 1, the illustrated node for illustrative purposes only , As SAN 100 can connect any number and any type of independent processor nodes, I / O fitness Adapter nodes and I / O device nodes. Any one node as a terminal node, which is fixed on this Defined as the origin or final consumption of messages in SAN 100 or frame of the device. ...
SAN 100 is a high-bandwidth, low-latency event network, which interconnects the distributed computer system Node. A node is attached to one or more of the network links in the network and a message is formed The starting and / or destination of any parts. In the example described, SAN 100 includes the following form Node: node main processor 102, a main processor node 104, redundant array of independent disk (RAID) Subsystem node 106 and I / O backplane node 108. In Figure 1, the illustrated node for illustrative purposes only , As SAN 100 can connect any number and any type of independent processor nodes, I / O fitness Adapter nodes and I / O device nodes. Any one node as a terminal node, which is fixed on this Defined as the origin or final consumption of messages in SAN 100 or frame of the device. ...
The message used in this application is limited to the data exchange unit, which is in cooperative communication between processes The letter of the original unit. Grouping is through networking protocol header and / or trailer, a unit of encapsulated data. Generally provide for the header frame via SAN boot control and routing information. Generally includes control and tail Cyclic redundancy check (CRC) data, used to ensure no damage to the contents of the packet provided.
The message used in this application is limited to the data exchange unit, which is in cooperative communication between processes The letter of the original unit. Grouping is through networking protocol header and / or trailer, a unit of encapsulated data. Generally provide for the header frame via SAN boot control and routing information. Generally includes control and tail Cyclic redundancy check (CRC) data, used to ensure no damage to the contents of the packet provided....
SAN 100 in Figure 1 includes a switch 112, switch 114, switch 146 and the router 117. Open Clearance is such a device, it is a plurality of links connected together and allows the use of a small object header To local identifiers (DLID) field from the subnet a link to another link transmitted packet. A router is a device like this, it will be multiple subnets connected together, and the ability to use a large header Destination globally unique identifier (DGUID) come from the first subnet a link to in the second sub- Pass another link within the frame.
In one embodiment, the link to the node such as terminals, switch, or any two network routers Contact between the structural elements of the duplex channels. As an example of suitable links include, but are not limited to, copper cables, optical Cables and the base plate and the printed circuit board printed circuit copper traces.
For reliable service types, such as a host processor end nodes and I / O adapter terminal node Terminal node generates and returns a confirmation packet request packet. Switches and routers transfer points from the source to the destination Groups. In addition to the updated each network-level CRC tail field changes, the switching of the packet no change Variable ground delivery. When the router passes the packet CRC updated when changing the suffix header field and modify its His fields.
In the SAN 100 shown in Figure 1, the host processor node 102, host processor node 104 and I / O Plate 108 includes at least one connection to the SAN 100 in Channel Adapter (CA). In one embodiment, In each channel adapter is an endpoint that achieve sufficient detail in the originating channel adapter interface Or absorbent structure 100 on the SAN transmitted packet. The host processor node 102 includes a main channel adapter 118 and the main channel adapter 120 in the form of a channel adapter. The host processor node 104 includes a main channel Adapter 122 and the main channel adapter 124. The main processor node 102 also includes a bus system 134 126-130 interconnected central processing unit and a memory 132. The main processor node 104 similarly includes Bus system 144 interconnects a central processing unit and a memory 142 136-140. ...
In the SAN 100 shown in Figure 1, the host processor node 102, host processor node 104 and I / O Plate 108 includes at least one connection to the SAN 100 in Channel Adapter (CA). In one embodiment, In each channel adapter is an endpoint that achieve sufficient detail in the originating channel adapter interface Or absorbent structure 100 on the SAN transmitted packet. The host processor node 102 includes a main channel adapter 118 and the main channel adapter 120 in the form of a channel adapter. The host processor node 104 includes a main channel Adapter 122 and the main channel adapter 124. The main processor node 102 also includes a bus system 134 126-130 interconnected central processing unit and a memory 132. The main processor node 104 similarly includes Bus system 144 interconnects a central processing unit and a memory 142 136-140. ...
In one embodiment, implemented in hardware main channel adapter. In this implementation, the main Channel adapter hardware offload the central processing unit and the I / O adapter most of the communication overhead. Main channel This hardware implementation adapter also allows the network through the switch and without simultaneous communication with a plurality of through- Communications protocol associated with the conventional overhead.
In one embodiment, implemented in hardware main channel adapter. In this implementation, the main Channel adapter hardware offload the central processing unit and the I / O adapter most of the communication overhead. Main channel This hardware implementation adapter also allows the network through the switch and without simultaneous communication with a plurality of through- Communications protocol associated with the conventional overhead....
Figure 1 of the I / O chassis 108 includes an I / O switch 146 and a plurality of I / O modules 148-156. In this Some examples, I / O module using the adapter card form. In the example shown in Figure 1 the adapter card comprising: The I / O module 148 SCSI adapter card; For for I / O module 152, Fibre Channel hub And a Fibre Channel arbitrated loop (FC-AL) device adapter card; for I / O module 150 to Ethernet adapter card; for I / O module 154 graphics adapter card; for I / O module 156 videos Adapter cards. Can achieve any known type of adapter card. I / O adapter are included in the I / O adapter Backplane of the switch for the adapter card is connected to the SAN fabric. These modules include target channel Adapter 158-166. ...
Figure 1 of the I / O chassis 108 includes an I / O switch 146 and a plurality of I / O modules 148-156. In this Some examples, I / O module using the adapter card form. In the example shown in Figure 1 the adapter card comprising: The I / O module 148 SCSI adapter card; For for I / O module 152, Fibre Channel hub And a Fibre Channel arbitrated loop (FC-AL) device adapter card; for I / O module 150 to Ethernet adapter card; for I / O module 154 graphics adapter card; for I / O module 156 videos Adapter cards. Can achieve any known type of adapter card. I / O adapter are included in the I / O adapter Backplane of the switch for the adapter card is connected to the SAN fabric. These modules include target channel Adapter 158-166. ...
SAN 100 processing for I / O and data communication between the processors. SAN 100 supports I / O required To the high bandwidth and scalability, and is required to support communications between the processor extremely low latency Time and extremely low CPU overhead. Users can bypass the operating system kernel customers and directly access Q network communication hardware, such as the main channel adapters that can achieve effective messaging protocol. SAN 100 is adapted to the current model, and for new forms of I / O and computer cluster communication in progress Li blocks. Moreover, the SAN 100 in Figure 1 such that I / O adapter node communication, or in between the In a distributed computer system of any or all of processor nodes. SAN 100 is attached to the I / O adapter case, I / O adapter node has in the SAN 100 in any of the main processor section Point substantially the same communication capabilities. ...
SAN 100 processing for I / O and data communication between the processors. SAN 100 supports I / O required To the high bandwidth and scalability, and is required to support communications between the processor extremely low latency Time and extremely low CPU overhead. Users can bypass the operating system kernel customers and directly access Q network communication hardware, such as the main channel adapters that can achieve effective messaging protocol. SAN 100 is adapted to the current model, and for new forms of I / O and computer cluster communication in progress Li blocks. Moreover, the SAN 100 in Figure 1 such that I / O adapter node communication, or in between the In a distributed computer system of any or all of processor nodes. SAN 100 is attached to the I / O adapter case, I / O adapter node has in the SAN 100 in any of the main processor section Point substantially the same communication capabilities. ...
In memory semantics, the source process to directly read or write to a remote node to the destination process's virtual Address space. Process requires only the remote destination communication data buffer position, and not Involves the transmission of any data. Thus, the semantics of the memory, the process of sending includes the destination source Process the destination address of the data packet buffer memory. Semantics in the memory, the destination process mentioned Source process before permission to access its memory.
Semantics and semantic memory channel typically I / O and interprocessor communications needs. Typical I / O channels, and a memory operation using a combination of semantics. In the illustrative example shown in Figure 1 a distributed computing Machine system I / O operations, such as a main processor node 102, host processor node using channel Semantic to start I / O operations, such as the RAID subsystem to the target channel adapter (TCA) 172 of the disc I / O adapter sends disk write commands. Disk I / O adapters examination orders and use direct memory semantics Node from the host processor reads the data buffer memory space. After the read data buffer, the disk I / O adapters use channel semantics pushed back to the main processor node I / O completion message. ...
Semantics and semantic memory channel typically I / O and interprocessor communications needs. Typical I / O channels, and a memory operation using a combination of semantics. In the illustrative example shown in Figure 1 a distributed computing Machine system I / O operations, such as a main processor node 102, host processor node using channel Semantic to start I / O operations, such as the RAID subsystem to the target channel adapter (TCA) 172 of the disc I / O adapter sends disk write commands. Disk I / O adapters examination orders and use direct memory semantics Node from the host processor reads the data buffer memory space. After the read data buffer, the disk I / O adapters use channel semantics pushed back to the main processor node I / O completion message. ...
Then turning to Figure 2, which describes the present invention according to a preferred embodiment the main processor section Point a functional block diagram. Such as a host processor node 200 in Figure 1 the primary host processor node 102 Examples of processor nodes.
Then turning to Figure 2, which describes the present invention according to a preferred embodiment the main processor section Point a functional block diagram. Such as a host processor node 200 in Figure 1 the primary host processor node 102 Examples of processor nodes....
202-208 consumer via the verbs interface 222 and message and data service 224 to transmit SAN Messages. Verb Interface is essentially the function of the main channel adapter explanatory memorandum. Operating system can pass Through its programming interface to reveal some or all of the verb functions. Generally, this interface defines the host's OK To. In addition, the host processor node 200 includes a message and data service 224, which is higher than the verb layer Level interface, and for processing by the channel adapter 210 and channel adapter 212 receives a message And data. Message and data service 224 provides an interface to consumers 202-208 to process messages and other Data.
Referring now to Figures 3A, according to the present invention, which describes a preferred embodiment of the main channel adapter The diagram shown in Figure 3A themain channel adapter 300A includes a plurality of queue pair (QP) 302A-310A, They are used for the mainchannel adapter port 312A-316A delivers the message. To the primarychannel adapter port 312A-316A of the data buffer by a virtual link (VL) 318A-334A boot, wherein each VL has its own flow control. Subnet manager configuration of the channel adapter port for each physical port that is The LID local address.
Sub-Manager Agent (SMA) 336A is the subnet manager for configuring the communication channel adaptation The entity. Memory translation and protection (MTP) 338A is the virtual address to physical address translation and And allows access to effective mechanisms. Direct Memory Access (DMA) 340A relative to the queue A memory for 302A-310A 340A to provide direct memory access operations.
As shown in Figure 3A of themain channel adapter 300A single channel adapter can support tens Wan queue right. In contrast, the I / O adapter of the target channel adapter typically supports a much smaller teams Column right. Each queue pair includes sending work queue (SWQ) and receive work queue. Send Task Force Column is used to send the message channel and memory semantics. Receive work queue receiving channel and semantic memory consumption Interest. Consumers call this operating system is called the verb programming interface specified in the work request (WR) into a work queue.
Figure 3B according to the present invention is described in a preferred embodiment of theswitch 300B.Switch 300B package Including thepacket relay device 302B, 306B which virtual channel such as by such a virtual channel and a plurality ofports 304B communicate. Typically, such a switch as theswitch 300B can be the same from one end of the switch Port to another port transmits data packets.
Similarly, FIG 3C described according to the present invention, one preferred embodiment of therouter 300C. Road Including grouping by therepeater unit 300C 302C, 306C it through virtual channels such as virtual channels andMultiple ports 304C communication. Like theswitch 300B asrouters 300C generally able in the same way from Router on a port to any other port transmits data packets.
Channel adapters, switches and routers use within a single physical link multiple virtual channels. As Figures 3A, 3B and 3C, the physical port of the terminal nodes, switches and routers to connect to the subnet. SAN fabric into the source of the packet from the packet, along with one or more virtual channels to the packet's Head Land. The selected virtual channel is associated with the packet service-level mapping. At any time, Only one virtual channel in a given serving as a physical link. Virtual channel provides a technique, For applying to a virtual channel link-level flow control rather than images of other virtual channels. When a Since the packets on the virtual channel contention, quality of service (QoS) or other factors obstruction, without On the same virtual channel as the packet is allowed. ...
Channel adapters, switches and routers use within a single physical link multiple virtual channels. As Figures 3A, 3B and 3C, the physical port of the terminal nodes, switches and routers to connect to the subnet. SAN fabric into the source of the packet from the packet, along with one or more virtual channels to the packet's Head Land. The selected virtual channel is associated with the packet service-level mapping. At any time, Only one virtual channel in a given serving as a physical link. Virtual channel provides a technique, For applying to a virtual channel link-level flow control rather than images of other virtual channels. When a Since the packets on the virtual channel contention, quality of service (QoS) or other factors obstruction, without On the same virtual channel as the packet is allowed. ...
Virtual channels provide deadlock avoidance. Virtual channels allow including circuit layout, in order to adopt all The physical link transmission packets, and to ensure the circuit is not likely to cause a deadlock caused by back pressure Lai Xing (back pressure dependency).
Virtual channel to reduce blocking of the line head (head-of-line blocking). When a switch is no longer There can be used for a given virtual channel credit packet, the use of different credit sufficient Virtual channels serving as a packet is allowed to continue.
Referring now to Figure 4, which describes the present invention according to a preferred embodiment of the graphic work, please Find the process in Fig. In Figure 4, there is a receivingwork queue 400, sendwork queue 402 and End Into thequeue 404 for processing to and from theconsumer 406 requests. These come from consumers, please Request is sent to thehardware 408 eventually. In this example, theconsumer 410 and 406 work requests generated 412, 414 and receives completed. Shown in Figure 4, is arranged on the work queue work requests Known work queue element (WQE).
Sendwork queue 402 includes a work queue element (WQE) 422-428, is used to describe the SAN Structure of data to be transmitted. Receivework queue 400 includes a work queue element (WQE) 416-420, From where the arrangement is used to describe the structure of the input channel SAN semantic data. Work queue element is in Main channel adapter hardware in 408 treatment.
Verb also provides a mechanism for retrieval from thecompletion queue 404 is done. Figure 4 Shows,completion queue 404 includes completion queue element (CQE) 430-436. Completion queue elements include On previously completed work queue element.Completion queue 404 is used for creating multiple queues for the complete A single point of completion notification. Completion queue element is accomplished in the queue data structure. This element describes The completed work queue element. The completion queue element include sufficient information to determine the queue pair and Complete specific work queue element. Done Queue range includes completion queue pointer pointing independent, Independently queue length and queue management independently additional information needed information blocks. ...
Verb also provides a mechanism for retrieval from thecompletion queue 404 is done. Figure 4 Shows,completion queue 404 includes completion queue element (CQE) 430-436. Completion queue elements include On previously completed work queue element.Completion queue 404 is used for creating multiple queues for the complete A single point of completion notification. Completion queue element is accomplished in the queue data structure. This element describes The completed work queue element. The completion queue element include sufficient information to determine the queue pair and Complete specific work queue element. Done Queue range includes completion queue pointer pointing independent, Independently queue length and queue management independently additional information needed information blocks. ...
A remote direct memory access (RDMA) read work request provides semantic memory operation On the remote node to read a virtual continuous memory space. Memory or a memory space can be Prefetching a part or a part of the memory window. Memory area referenced through virtual addresses and Length defined previously registered a group of virtual memory addresses continuously. Memory window reference has been Tied to the previously registered in the region of a group of virtual sequential memory addresses.
A remote direct memory access (RDMA) read work request provides semantic memory operation On the remote node to read a virtual continuous memory space. Memory or a memory space can be Prefetching a part or a part of the memory window. Memory area referenced through virtual addresses and Length defined previously registered a group of virtual memory addresses continuously. Memory window reference has been Tied to the previously registered in the region of a group of virtual sequential memory addresses....
RDMA write work queue element provides semantic memory operation on the remote node to write a Virtual continuous memory space. RDMA write work queue elements include local virtual continuous memory space And one of a write to the local memory space of the remote memory space distribution list of virtual address.
RDMA FetchOp work queue element provides semantic memory operation to the remote execution of the original vocabulary Sub-operation (atomic opertion). RDMA FetchOp work queue element is a combination of RDMA Read Access, modify, and RDMA write operations. RDMA FetchOp work queue element can support several read Take - modify - write operations, such as if the same time, such a comparison and exchange.
Binding (bind) (not bundled) Remote Access Key (R_Key) work queue element to the main channel Adapter hardware provides a command to the memory via the memory window region associated (or disassociated) To modify (destroy) memory window. R_Key RDMA access of each part and for Verify that the remote process has permission to access on the buffer.
In one embodiment, the receiver shown in Figure 4 thework queue 400 only supports a work queue Element, it is called to receive work queue element. Receive work queue element provides a channel semantics Operation, which is used to describe the transmission of messages written to the input space of the local memory. Receive work queue Elements include a description of several virtual memory space contiguous distribution lists. Input transmission message is written Into the memory space. Virtual address in the process of establishing a local queue for the address range.
For the communication between processors, user-mode software processes through the queue directly from the buffer in Resides in memory to transfer data. In one embodiment, the transfer through the queue pairs bypasses Operating system and consumes less primary instruction cycle. Copy queue of the processor data transfer allows zero, Which does not involve any of the core operating system. Zero-copy data transfer processor provide high bandwidth and low latency Communication time effective support.
When you create a queue for when the queue is set to provide more choices for the type of delivery service. In one embodiment, the present invention is a distributed computer system supports four types of delivery services: reliable , Unreliable, reliable and unreliable datagram datagram connection services.
Reliable and unreliable connection services on a local queue with one and only one remote On the associated queue. Connect service requires a process to establish communication used to pass each of the SAN fabric A process in a queue pair. Therefore, if the N primary processor nodes each of which includes P number of processes, And on each node of a process of hope for all P and all other nodes in all processes through Letter, each host processor node requires P2x (N-1) queues right. Moreover, a process can be A queue connected to the same main channel adapter on another queue right.
In Figure 5 generally illustrates the use of a reliable connection service to communicate between processes in a distributed Distributed computer system. The distribution in Figure 5 thecomputer system 500 includes a mainprocessor section Point 1, the main processor and the main processor node 2 Node 3. 1 includes a host processor node processes A510. Host processor node 2 and processes, including processes C520 D530. The main processor node 3, includingprocess E 540.
Node 1 includes a main processor queue for 4, 6 and 7, each having to send work queue and receive work For the queue. The main processor node 2 has a queue for 9, host processor node 3 has a queue pairs 2 and 5. Reliable distributedcomputer system 500 service will connect with a local queue and only one on the far Associated queue right away. Therefore, the queue and the queue for the four pairs of two for communication; queue with a team of seven for Column 5 Communication; queue for six of nine for communication with the queue.
Is arranged in a reliable connection service on WQE queue for making the data to be written by Lane Then the reception queue for receiving WQE referenced memory space. RDMA operations on the connected queue On the role of the address space.
In one embodiment of the present invention, making the reliable connection service, reliable, because the hardware to keep the serial number, And make sure all of the packet transmission. Hardware and software combinations SAN drives retry any failed Communication. Customers queue for the process even in the presence of bit errors, the receiver underrun (underrun) And the case of network congestion for reliable communication. If there is a SAN alternative road network Trail, you can switch even in the presence of structure, link or channel adapter port can also be the case of failure Enough to maintain reliable communication.
In addition, you can use the SAN structure confirmed to reliably deliver data. The confirmation is also available Not the process level to confirm that the process of validation data confirm the data has been consumed. Alternatively, recognized So the only indication data has reached its destination confirmation.
Reliable datagram service the local end (EE) and only a scope and a remote terminal To the end of the range associated with it. Unreliable datagram service allows a queue on the client process and in any Other remote nodes communicate with any other queue pairs. In the receiving queue, unreliable datagram service Service allow connections from any other remote node in any of the input message to send work queue.
Reliable datagram service greatly improved scalability, because reliable datagram service is no Connection. Thus, a fixed number of queue pairs of terminal nodes can be connected to a reliable transfer over Delivery service processes and more terminal nodes, with reliable datagram service through the process and end nodes Letter. For example, if N primary processor nodes each of which includes a process P, and on each node I hope all of P processes on all other nodes in all the process of communication, then the reliable connectivity services Service needed on each node P2x (N-1) queues right. By comparison, connectionless unreliable datagram service Service on each node need only cohort of P + (N-1) EE range for the same traffic.
Figure 6 illustrates the use of a reliable datagram service to communicate between processes in a distributed distributed meter Computer system. Figure 6 distributedcomputer system 600 includes amain processor node 1, the main processor section Point 2 and the main processor node 3. 1 comprises a main processor node processes the queue for four A610. Main Processor node 2 includes a queue process for 24 C620 and the process of having a queue for 25 D630. Node 3 includes a host processor 14 processes the queue for E640.
In a distributedcomputer system 600 to achieve a reliable datagram service, the queue for the connection in the So-called connectionless transport services within. For example, a reliable datagram service to connect to the queue for four queues on 24, 25 and 14. Specifically, unreliable datagram service allows the transmission queue for four reliably work queue To the 24, 25 and 14 in the queue on the receiving work queue delivers the message. Similarly, the queue for the 24, 25 and 14 can reliably transmit queue in the queue for 4 to receive work queue delivers the message.
In one embodiment of the present invention, the use of reliable datagram service is associated with each message frame The serial number and validation services to ensure a reliable connection with the same degree of reliability. End (EE) Fan Maintaining end-around to record the serial number specific status, recognition and timeout values. Stored within the EE End state is between a pair of terminal nodes of all non-connection queue for communications sharing. Each terminal EE node requires at least one range to be used in unreliable datagram it wishes to communicate with each service End nodes (e.g., a given terminal node to the other terminal to the N nodes have Reliable datagram service of at least N EE range). ...
In one embodiment of the present invention, the use of reliable datagram service is associated with each message frame The serial number and validation services to ensure a reliable connection with the same degree of reliability. End (EE) Fan Maintaining end-around to record the serial number specific status, recognition and timeout values. Stored within the EE End state is between a pair of terminal nodes of all non-connection queue for communications sharing. Each terminal EE node requires at least one range to be used in unreliable datagram it wishes to communicate with each service End nodes (e.g., a given terminal node to the other terminal to the N nodes have Reliable datagram service of at least N EE range). ...
Then turning to Figure 7, which describes the present invention according to a preferred embodiment of a data packet Icon. A data packet information passed through the SAN fabric units. Data packet is the terminal section The structure point to the terminal node, so it is the terminal node creation and consumption. To reach the channel adapter for (Master or target) packet, the data packet is not in the SAN fabric switches and routers produced or consumed. With respect to the packet reaches the opposite channel adapters, switches and routers merely request packet or OK Mobile packet closer to the final destination, the process used to modify the variable link header fields. Road When a packet through the router subnet boundaries also modify the network packet header. After passing through the subnet, a single sub- Group reside on a single service level. ...
Message data 700 includes data section 1702, the data segment 2704 and data segment 3706, which class Illustrated in Figure 4 similar data segment. In this example, these data segments form apacket 708, which is 712 disposed to the data packet payload of thepacket 710. In addition, thedata packet 712 including theCRC 714, it is used for error detection. In addition, thedata packet 712 exists in therouting header 716 andpass Send 718.Routing header 716 is used to identify the source of the data packet and thedestination port 712. In this An example of the transmissiondata packet header 712 718 specifies the destination queue pair. In addition, thefirst transmission Standard 718 is also provided information, such as a data packet of theoperation code 712, the packet sequence number, and partition. ...
Message data 700 includes data section 1702, the data segment 2704 and data segment 3706, which class Illustrated in Figure 4 similar data segment. In this example, these data segments form apacket 708, which is 712 disposed to the data packet payload of thepacket 710. In addition, thedata packet 712 including theCRC 714, it is used for error detection. In addition, thedata packet 712 exists in therouting header 716 andpass Send 718.Routing header 716 is used to identify the source of the data packet and thedestination port 712. In this An example of the transmissiondata packet header 712 718 specifies the destination queue pair. In addition, thefirst transmission Standard 718 is also provided information, such as a data packet of theoperation code 712, the packet sequence number, and partition. ...
In Figure 8, it is described as part of a distributed computer system with a graphic example of the request and validation Given transaction. Figure 8 is a distributed computer system includes amain processor node 802 and nodemain processor 804. Thehost processor node 802 includesprimary channel adapter 806. Thehost processor node 804 includes a mainchannel Channel adapter 808. Figure 8 is a distributed computer system includes aSAN fabric 810, which includes aswitch 812 andswitch 814. SAN fabric including themain channel adapter 806 connected to theswitch 812 of the link, Theswitch 812 is connected to theswitch 814 of the link, theprimary channel adapter 808 connected to theswitch 814 Links.
% E5% 9C% A8% E7% A4% BA% E4% BE% 8B% E4% BA% 8B% E5% 8A% A1% E4% B8% AD% EF% BC% 8C% E4% B8% BB% E5 % A4% 84% E7% 90% 86% E5% 99% A8% E8% 8A% 82% E7% 82% B9802% E5% 8C% 85% E6% 8B% AC% E5% AE% A2% E6% 88 % B7% E7% AB% AF% E8% BF% 9B% E7% A8% 8BA% E3% 80% 82% E4% B8% BB% E5% A4% 84% E7% 90% 86% E5% 99% A8 % E8% 8A% 82% E7% 82% B9804% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 0A% 09% 09% 09% 09% 09 % E5% 8C% 85% E6% 8B% AC% E5% AE% A2% E6% 88% B7% E7% AB% AF% E8% BF% 9B% E7% A8% 8BB% E3% 80% 82% E5 % AE% A2% E6% 88% B7% E7% AB% AF% E8% BF% 9B% E7% A8% 8BA% E9% 80% 9A% E8% BF% 87% E9% 98% 9F% E5% 88 % 97% E5% AF% B9824% E4% B8% 8E% E4% B8% BB% E4% BF% A1% E9% 81% 93% E9% 80% 82% E9% 85% 8D% E5% 99% A8 % E7% A1% AC% E4% BB% B6806% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% E4% BA% A4% E4% BA% 92 % E3% 80% 82% E5% AE% A2% E6% 88% B7% E7% AB% AF% E8% BF% 9B% E7% A8% 8BB% E9% 80% 9A% E8% BF% 87% E9 % 98% 9F% E5% 88% 97% E5% AF% B9828% E4% B8% 8E% E7% A1% AC% E4% BB% B6% E4% B8% BB% E4% BF% A1% E9% 81 % 93% E9% 80% 82% E9% 85% 8D% E5% 99% A8808% E4% BA% A4% E4% BA% 92% E3% 80% 82% E9% 98% 9F% E5% 88% 97 % E5% AF% B9% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20824% E5% 92% 8C828% E6% 98% AF% E5% 8C% 85 % E6% 8B% AC% E5% 8F% 91% E9% 80% 81% E5% B7% A5% E4% BD% 9C% E9% 98% 9F% E5% 88% 97% E5% 92% 8C% E6 % 8E% A5% E6% 94% B6% E5% B7% A5% E4% BD% 9C% E9% 98% 9F% E5% 88% 97% E7% 9A% 84% E6% 95% B0% E6% 8D % AE% E7% BB% 93% E6% 9E% 84% E3% 80% 82
A process through to the queue on the sending queue 824 registered work queue element to start the message please Requirements. Figure 4 illustrates such a work queue element. A client process sends a message request is in Work queue element includes a summary list of references. In summary list of each data point to the virtual segment Continuous local memory area, which includes part of the message, as shown by the data segments 2, 3 Those, which are stored in the message portion of Figure 4, 2, 3.
In themain channel adapter 806 hardware to read the work queue element, and will be stored in the virtual Lian Continued the message buffer is divided into data packets, such as the data packet shown in Figure 7. Data packets through Passed over the SAN fabric, and for reliable delivery service is the final destination terminal node to confirm. As If that was not successfully recognized, the data packet transmitted by the source terminal node again. Data packet by the source terminal Node generated and consumed by the destination terminal node.
Referring to Figure 9, described according to the invention has been described in a distributed network system using a network Addressing FIG. Providing primary name as the main processor for nodes or I / O adapter node of the master node Logical identifier. Main terminal node name identifies the message so that the message arrives at the end of the specified by the host name Processes residing on the client node. Thus, each electromechanical one primary name, but a node With multiple CA.
An individual identifier assigned by the IEEE 64-bit (EUI-64) 902 is assigned to each component. A One component may be a switch, router, or CA.
EachCA port 906 assigned one or more globally unique ID (GUID)identifier 904. By the For a number of reasons the use of a plurality of GUID (also called an IP address), some of which are the following examples illustration. In one embodiment, a different IP address identifying a terminal node on a different partition or services. In various embodiments, using a different IP address to specify a different quality of service (QoS) attributes. In another embodiment, a different IP subnet address recognition by different routing paths. AGUID 908 is assigned to theswitch 910.
Local ID (LID) refers to a single subnet used to identify the CA port short address ID. In one example embodiment, a subnet with up to 216Local ID (LID) refers to a single subnet used to identify the CA port short address ID. In one example embodiment, a subnet with up to 2...LMCLocal ID (LID) refers to a single subnet used to identify the CA port short address ID. In one example embodiment, a subnet with up to 2...
LID and GUID do not have a one to one correspondence between, because the CA can have more than each GUID of the port more or less LID. For a multiple redundant SAN fabric ports and redundant I conductive CA, CA at each of its endpoints may but are not required to use the same LID and GUID.
% E5% 9B% BE10% E4% B8% AD% E5% 9B% BE% E8% A7% A3% E4% BA% 86% E6% 8C% 89% E7% 85% A7% E6% 9C% AC% E5 % 8F% 91% E6% 98% 8E% E7% 9A% 84% E4% B8% 80% E4% B8% AA% E4% BC% 98% E9% 80% 89% E5% AE% 9E% E6% 96 % BD% E4% BE% 8B% E7% 9A% 84% E5% 88% 86% E5% B8% 83% E5% BC% 8F% E8% AE% A1% E7% AE% 97% E6% 9C% BA % E7% B3% BB% E7% BB% 9F% E7% 9A% 84% E4% B8% 80% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20 % E9% 83% A8% E5% 88% 86% E3% 80% 82% E5% 88% 86% E5% B8% 83% E5% BC% 8F% E8% AE% A1% E7% AE% 97% E6 % 9C% BA% E7% B3% BB% E7% BB% 9F1000% E5% 8C% 85% E6% 8B% AC% E5% AD% 90% E7% BD% 911002% E5% 92% 8C% E5% AD % 90% E7% BD% 911004% E3% 80% 82% E5% AD% 90% E7% BD% 911002% E5% 8C% 85% E6% 8B% AC% 0A% 20% 20% 20% 20% 20 % 20% 20% 20% 20% 20% 20% 20% E4% B8% BB% E5% A4% 84% E7% 90% 86% E5% 99% A8% E8% 8A% 82% E7% 82% B91006 % E3% 80% 811008% E5% 92% 8C1010% E3% 80% 82% E5% AD% 90% E7% BD% 911004% E5% 8C% 85% E6% 8B% AC% E4% B8% BB% E5 % A4% 84% E7% 90% 86% E5% 99% A8% E8% 8A% 82% E7% 82% B91012% E5% 92% 8C1014% E3% 80% 82% 0A% 20% 20% 20% 20 % 20% 20% 20% 20% 20% 20% 20% 20% E5% AD% 90% E7% BD% 911004% E5% 8C% 85% E6% 8B% AC% E4% B8% BB% E5% A4 % 84% E7% 90% 86% E5% 99% A8% E8% 8A% 82% E7% 82% B91012% E5% 92% 8C1014% E3% 80% 82% E5% AD% 90% E7% BD% 911002 % E5% 8C% 85% E6% 8B% AC% E5% BC% 80% E5% 85% B31016% E5% 92% 8C1018% E3% 80% 82% 0A% 20% 20% 20% 20% 20% 20 % 20% 20% 20% 20% 20% 20% E5% AD% 90% E7% BD% 911004% E5% 8C% 85% E6% 8B% AC% E5% BC% 80% E5% 85% B31020% E5 % 92% 8C1022% E3% 80% 82
Router subnet. For example, the subnet router 1022 1024 and 1026 used to connect to the sub- Network 1004. In one exemplary embodiment, the sub-terminal nodes with up to 216, switching and routing Makers.
Subnet is defined as a single unit is managed as a group of terminal nodes and cascaded switches. Typically, the sub- Network occupies a single geographical or functional areas. For example, in a room can be a single computer system Defined as a subnet. In one embodiment, a switch in a subnet can be executed very quickly, for Message type or plunge wormhole routing (wormho1e or cut-through routing).
Subnet within a subnet switch test only DLID, to allow the customers to quickly and efficiently Pass input message packet. In one embodiment, the switch is a relatively simple circuit, and Is typically implemented as a single integrated circuit. Subnet can have thousands of customers formed by the end of the cascade End node.
Shown in Figure 10, in order to expand into a larger system, such as a router 1024 and 1026 using the Router to connect subnets. Translation router IP destination ID (for example IPv6 destination ID), and Pass IP class groupings.
An exemplary embodiment of the switch is generally illustrates in Figure 3B. In each of the switch or router An I / O path having a port. Typically, a switch can be on the same switch from one port Transfer to any other port grouping.
In 1002, such as subnet or subnets 1004 a subnet from source port to destination port A path is the destination host channel adapter port LID OK. Between subnets by destination Main channel adapter port IP address (e.g. IPv6 address) and will be used to reach the destination by the sub The other end of the network port address to determine the path of LID.
In one embodiment, the request packet is not required to use a path and a corresponding request packet is certainly Acknowledgment (ACK) or negative acknowledgment (NAK) frame symmetry. Forget the use of an embodiment of a route , The switch to select the output port according DLID. In one embodiment, the switch uses a set of routing Criteria to serve all of the input ports. In one example embodiment, in a routing table. Including routing criteria. In one alternative embodiment, the switch is used to separate a set of standards for Each input port. In the present invention, a distributed computer system typically includes several hardware data transaction And software steps. Client process data transmission services can be user-mode or kernel-mode process. Client process through, for example in Figure 3A, 5 and 6 are illustrated in the queue for one or more queue pairs Visit the main channel adapter hardware. Client process calls the operating system-specific programming interface that this was Called a "verb." Software code to achieve the verb given work queue queue for registration work queue element Su. ...
In one embodiment, the request packet is not required to use a path and a corresponding request packet is certainly Acknowledgment (ACK) or negative acknowledgment (NAK) frame symmetry. Forget the use of an embodiment of a route , The switch to select the output port according DLID. In one embodiment, the switch uses a set of routing Criteria to serve all of the input ports. In one example embodiment, in a routing table. Including routing criteria. In one alternative embodiment, the switch is used to separate a set of standards for Each input port. In the present invention, a distributed computer system typically includes several hardware data transaction And software steps. Client process data transmission services can be user-mode or kernel-mode process. Client process through, for example in Figure 3A, 5 and 6 are illustrated in the queue for one or more queue pairs Visit the main channel adapter hardware. Client process calls the operating system-specific programming interface that this was Called a "verb." Software code to achieve the verb given work queue queue for registration work queue element Su. ...
In one embodiment, the channel adapter hardware testing work queue element registration and access to the labor For the queue element. In this embodiment, the channel adapter hardware translation and validation work queue element Virtual address and access the data.
In one embodiment, the channel adapter hardware testing work queue element registration and access to the labor For the queue element. In this embodiment, the channel adapter hardware translation and validation work queue element Virtual address and access the data....
If reliable delivery service, the data packet when it reaches its destination end node, Used by the destination terminal results confirm the data packet to make a request packet sender knows the requested data Packets at the destination is validated and accepted. Acknowledgment data packet confirm one or more valid and accepted The request data packet. Requester can be recognized before it receives any of a plurality of pending requests Packets. In one embodiment, when establishing a queue of pending messages when determining a plurality, i.e. Requested number of data packets.
Used in the present invention, the hierarchical structure of an embodiment 1100 in Figure 11 in the form of FIG. General illustration. Figure 11 illustrates the hierarchical structure of the data communication paths of the various layers and the layers Between data and control information through the organization.
Main channel adapter terminal node protocol layer (e.g., use by the terminal node 1111) comprises a consumption 1103 defined by the upper layer protocol 1102, the transport layer 1104, a network layer 1106, the link layer 1108 and was Physical layer 1110. Switching level (for example, used by switch 1113), including physical layer link layer 1108 and 1110. Router layer (for example, a router 1115 used) includes a network layer 1106, the link layer 1108 and the physical Layer 1110.
1100 generally follows the classic hierarchical communication stack contours. The agreement on the terminal node 1111 Yee layer, for example, the verb upper layer protocol 1102 (1112) to create messages in the transmission layer 1104. 1104 to the network transport layer transfer message 1106 (1114). In 1106 the network layer network subnet (1116) Passed between the groupings. 1108 link layer packet transmitted in the network subnet (1118). 1110 to the physical layer The physical layer to send other devices bits or bit groups. Each layer of the upper or lower do not know how to do it Their function.
1103 and 1105, consumption by using the node for communication between the terminal application of other layers Or processes. 1104 message transport layer provides end-to move. In one embodiment, the transfer layer provides such The transmission of the three types of services that are reliable connection service; reliable datagram service; Unreliable datagram service. 1106 executive through a network layer subnet or multiple subnets to the final destination End node routing of packets. Link Layer 1108 execution flow control, error checking and cross-link prioritization Packet transmission.
1110 physical layer implementation of technology-related bits transmitted. Bits or bit groups via link 1122 1124 and 1126 in the physical layer transmission. You can use printed circuit board copper traces, copper, fiber or With other appropriate links to implement the link.
As described above, the present invention relates to improvement of the system area network devices, systems and methods, the Said system area network such as InfiniBand network in which the logical partition is the primary channel system LAN Channel adapter (HCA) support. Using the present invention provides a support for the HCA logical partitions, and more Of operating systems can share a single physical HCA resources. Logical partitions to ensure that each operating system I do not know Road is being shared with other operating systems HCA hardware resources, and also to ensure the independence prevent specific partition Operating systems and applications to access other partitions associated with HCA hardware resources.
Figure 12 is a diagram illustrating the main channel adapter supports physical logic elements shown in the main channel adapter Intentions. Shown in Figure 12 corresponds to the main channel adapter illustrated in Figures 3A main channel adapter, but with With the present invention is to provide improvements in the primary channel adapter support the logical partitions.
Shown in Figure 12, the main channel adapter (HCA) 1200 shown in Figure 3A includes a main channel suitable Orchestration similar two physical ports 1210 and 1220. In addition, the HCA 1200 provided on each queue 1232-1242, is used to send data to the host system resources from the primary system resource group and receiving data packets. Additional facilities are provided, they allow the logic switches and logic HCA as the external subnet manager through Often switch and HCA appear.
The host system resources manner known in the art is logically partitioned into logical partitions LPAR1-LPARn. For each LPAR with the main channel adapter accurate view of the structure of the tube Create a logical processor adapter main channel image, wherein one or more logical ports attached to a logic Series switch, said logic switch in turn is connected to the main channel physical port adapter. Such as Action taken including the allocation of separate main channel adapter resources, such as queue pairs, completion queue, Event queue, the main channel for each logical adapter / logical port to send and receive data counter and storage Memory translation table entry. These assignments are not normally occur simultaneously, as a resource request from the application with The temporal dynamics. However, when each resource is allocated, the resource allocation manager has logic Partition logical partition identifier to verify ownership during operation. ...
Returning to Figure 12, some of the queues in the HCA to 1260 assigned by the manager logical partition The logical partition identifiers associated LPAR1-LPARn. For example, when a logical partition operating system Request to use than another to be assigned to another logical partition on a queue when the queue that will be Distribution corresponding to the operating system's logical partition logical partition identifier. In this manner, only the management Device (it is possible to access any HCA resources credible software), operating system and with the operating system logic Logical partitions associated user application can access the queue right. Moreover, the operating system and the logical partition should be With the inspection of the operating system and application-specific system, and does not know about the other logical partitions. ...
Returning to Figure 12, some of the queues in the HCA to 1260 assigned by the manager logical partition The logical partition identifiers associated LPAR1-LPARn. For example, when a logical partition operating system Request to use than another to be assigned to another logical partition on a queue when the queue that will be Distribution corresponding to the operating system's logical partition logical partition identifier. In this manner, only the management Device (it is possible to access any HCA resources credible software), operating system and with the operating system logic Logical partitions associated user application can access the queue right. Moreover, the operating system and the logical partition should be With the inspection of the operating system and application-specific system, and does not know about the other logical partitions. ...
Each logical partition has its own address space, in order to prevent unauthorized software access logic Partition hardware resources. Just as a trusted code and have access to hardware resources highest priority tube Processors, operating systems and logical partition in the logical partition applications can be assigned to the logical access points District HCA resources.
Each logical partition has its own address space, in order to prevent unauthorized software access logic Partition hardware resources. Just as a trusted code and have access to hardware resources highest priority tube Processors, operating systems and logical partition in the logical partition applications can be assigned to the logical access points District HCA resources....
In other words, each having an associated logical port LID / GID. Logical port belongs has been Assigned to a logical partition of the logical host channel adapter. Cohort of 1232-1242 and the specific logic Associated with the port. In this manner, each logical partition works like it has its own logic primary channel Channel adapter 1282-1286.
Each logical adapter 1282-1286 main channel may have been assigned to more than one logical partition Queues for 1232-1242, they have to be configured to provide the access multiple physical ports 1292-1294 Q. properties. Because each queue for 1232-1242 can be sent to multiple physical ports 1292-1294 Data packets and receive packets, so each queue for 1232-1242 to provide such a facility, they refer to Are shown associated with the physical port.
Each physical port 1292-1294 also provide additional facilities, which are used to control to a specific purpose Ground LID packet is being sent out or delivered to the physical port 1292-1294 insourcing in HCA 1200 Including the queue right. Namely, HCA hardware to test using port facilities to be sent to a particular queue for 1232-1242 inward grouping in order to ensure that they meet the queue on the property. HCA also uses hardware Port facilities to test out a packet to determine whether the destination is within the physical HCA 1200 another Queues for 1232-1242, in this case, the transfer packet to the internal queue of 1232-1242. HCA Hardware in the following ways to perform this test: View destination logical identifier (DLID) to see DLID are assigned to the master by the physical channel adapter 1200 supported logical host channel adapter Is defined on the logical port 1282-1286. These packets are transmitted to the inside in order to ensure that it is checked We also meet the destination queue on the property. Facilities and port facilities on the queue manager software is saved, And established in the queue and the physical ports on the switch between the logical view. ...
Each physical port 1292-1294 also provide additional facilities, which are used to control to a specific purpose Ground LID packet is being sent out or delivered to the physical port 1292-1294 insourcing in HCA 1200 Including the queue right. Namely, HCA hardware to test using port facilities to be sent to a particular queue for 1232-1242 inward grouping in order to ensure that they meet the queue on the property. HCA also uses hardware Port facilities to test out a packet to determine whether the destination is within the physical HCA 1200 another Queues for 1232-1242, in this case, the transfer packet to the internal queue of 1232-1242. HCA Hardware in the following ways to perform this test: View destination logical identifier (DLID) to see DLID are assigned to the master by the physical channel adapter 1200 supported logical host channel adapter Is defined on the logical port 1282-1286. These packets are transmitted to the inside in order to ensure that it is checked We also meet the destination queue on the property. Facilities and port facilities on the queue manager software is saved, And established in the queue and the physical ports on the switch between the logical view. ...
Manager 1260 provides a virtual address translation and protection table, enter the project with its unique HCA resources Li logical partition associated with it. That is, in each facility and the protection of the virtual address translation table with the An input item, and the input item can be associated with a particular logical partition. By the Said virtual address translation and protection tables provide separate logical partitions, and the underlying operating system queues Associated direct memory access (DMA) operation is limited to the authorization of that queue memory space. Thus, for example, in a logical partition of theapplication 1 can not access the resources of logical partitions 2 because the And protection of the virtual address translation tables logical partition resources address two logical partitions an irrelevant Union. In other words, the logical partition resources 2 in the logical partition address will exceed 1 virtual ground Address translation and protection table range. ...
Manager 1260 provides a virtual address translation and protection table, enter the project with its unique HCA resources Li logical partition associated with it. That is, in each facility and the protection of the virtual address translation table with the An input item, and the input item can be associated with a particular logical partition. By the Said virtual address translation and protection tables provide separate logical partitions, and the underlying operating system queues Associated direct memory access (DMA) operation is limited to the authorization of that queue memory space. Thus, for example, in a logical partition of theapplication 1 can not access the resources of logical partitions 2 because the And protection of the virtual address translation tables logical partition resources address two logical partitions an irrelevant Union. In other words, the logical partition resources 2 in the logical partition address will exceed 1 virtual ground Address translation and protection table range. ...
As mentioned above, the queue manager is responsible for the allocation of 1260 logical partition identifier. This logic points Zone identifier in the queue for HCA and their associated resources grouped into logical partitions. The logical partition Identifier is also used when the queue from a specific input on the work queue item (WQE) generated completion queue Enter the project (CQE) or event queue input items (EQE) when executed test.
As mentioned above, the queue manager is responsible for the allocation of 1260 logical partition identifier. This logic points Zone identifier in the queue for HCA and their associated resources grouped into logical partitions. The logical partition Identifier is also used when the queue from a specific input on the work queue item (WQE) generated completion queue Enter the project (CQE) or event queue input items (EQE) when executed test....
As mentioned above, the main channel adapter (HCA) typically includes support for multiple queues for (QP), End A queue (CQ), the event queue (EQ) and a memory area (MR) memory required facilities. In this use, the term "facility" refers to any hardware resources such as registers. Queue for commonly Assigned to run on the host processor user program. In addition to the queue for the groups, the completion queue, it Member queues, as described above, the memory area can be assigned to other logical partitions.
Expected to allow only a specific queue for associated program to access the HCA in the queue for the design Shi. Moreover, it is desirable to allow only be assigned to a logical partition of the operating system and user application access and Logical partitions that are associated HCA facilities. In this manner, a logical partition operating system and User application does not interfere with another logical partition operating system and user application functionality. Moreover, as If a logical partition of failure or malfunction, then the failure or malfunction of the logic to be associated with that particular Logical partition isolation, and does not interfere with other logical partitions. The present invention provides a mechanism, with To control the access to the facility for HCA, only the logical partition to user applications and operating system or various As manager trusted code can access a given logical partition HCA facilities. ...
Expected to allow only a specific queue for associated program to access the HCA in the queue for the design Shi. Moreover, it is desirable to allow only be assigned to a logical partition of the operating system and user application access and Logical partitions that are associated HCA facilities. In this manner, a logical partition operating system and User application does not interfere with another logical partition operating system and user application functionality. Moreover, as If a logical partition of failure or malfunction, then the failure or malfunction of the logic to be associated with that particular Logical partition isolation, and does not interfere with other logical partitions. The present invention provides a mechanism, with To control the access to the facility for HCA, only the logical partition to user applications and operating system or various As manager trusted code can access a given logical partition HCA facilities. ...
By the address translation mechanism to control access to the memory of the corresponding page in said address translation Mechanism, the virtual address is translated into system memory address space of the physical address. According to such applications, Operating system, trusted code such requestor's identity, said address translation mechanism may determine whether Allow the requester to access the system virtual address of the mapped memory pages. The logical identifier may be for example Partition identifier and so on. In this way, from one logical partition can not access the requester is assigned To a different logical partition system memory page.
Assumes the requestor access to the virtual address associated with the system memory page, corresponds to the virtual Address is the actual address of the page from the system memory are obtained, and for access is assigned to the team Column, the completion queue, event queue system memory or the memory area address space using the class page Surface. Using the actual address to access specific pages using the class using the class identify the requester. Moreover, the By using the class has a corresponding page includes the use of class requester access to a list of HCA facilities. Then compare request access facility class identifier and the use of the page identifier of the HCA facility. If the requested access facility class identifier matches the page using the identifier of the HCA facility, the Permit access to the facility for HCA, such as written or read. Otherwise, access is denied. ...
Assumes the requestor access to the virtual address associated with the system memory page, corresponds to the virtual Address is the actual address of the page from the system memory are obtained, and for access is assigned to the team Column, the completion queue, event queue system memory or the memory area address space using the class page Surface. Using the actual address to access specific pages using the class using the class identify the requester. Moreover, the By using the class has a corresponding page includes the use of class requester access to a list of HCA facilities. Then compare request access facility class identifier and the use of the page identifier of the HCA facility. If the requested access facility class identifier matches the page using the identifier of the HCA facility, the Permit access to the facility for HCA, such as written or read. Otherwise, access is denied. ...
Chartered kind of resource is to be protected against the user program, but by the logic associated with the resource Partition running an operating system and firmware access resources. The resources in this class is not so The user program can be accessed directly, and only a single logical partition can affect those resources.
% E5% AE% 9E% E9% 99% 85% E5% 9C% B0% E5% 9D% 80% E7% B1% BB% E8% B5% 84% E6% BA% 90% E6% 98% AF% E9 % 82% A3% E4% BA% 9B% E5% 8C% 85% E6% 8B% AC% E5% AE% 9E% E9% 99% 85% E5% 9C% B0% E5% 9D% 80% E7% 9A % 84HCA% E8% B5% 84% E6% BA% 90% E3% 80% 82% E5% AF% B9% E4% BA% 8E% E5% AE% 9E% E9% 99% 85% E5% 9C% B0 % E5% 9D% 80% E7% B1% BB% E8% B5% 84% E6% BA% 90% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20 % E8% A6% 81% E6% B1% 82% E4% B8% 8E% E7% 89% B9% E8% AE% B8% E7% B1% BB% E7% 9B% B8% E5% 90% 8C% E7 % 9A% 84% E8% AE% BF% E9% 97% AE% E8% A6% 81% E6% B1% 82% E3% 80% 82% E8% BF% 99% E4% B8% AA% E4% BD % BF% E7% 94% A8% E7% B1% BB% E7% 94% A8% E4% BA% 8E% E6% 94% AF% E6% 8C% 81% E8% 99% 9A% E6% 8B% 9F % E6% 9C% BA% E5% BA% 94% E7% 94% A8% E5% 92% 8C% E6% A0% 87% E5% BF% 97% E8% B5% 84% E6% BA% 90% EF % BC% 8C% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% E6% 89% 80% E8% BF% B0% E6% A0% 87% E5 % BF% 97% E8% B5% 84% E6% BA% 90% E5% 85% B7% E6% 9C% 89% E5% BF% 85% E9% A1% BB% E4% BB% 8E% E9% 80 % BB% E8% BE% 91% E5% 88% 86% E5% 8C% BA% E5% AE% 9E% E9% 99% 85% E5% 9C% B0% E5% 9D% 80% E7% BF% BB % E8% AF% 91% E4% B8% BA% E7% B3% BB% E7% BB% 9F% E5% AE% 9E% E9% 99% 85% E5% 9C% B0% E5% 9D% 80% E7 % 9A% 84% E5% AE% 9E% E9% 99% 85% E5% 9C% B0% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% E5 % 9D% 80% E3% 80% 82
Chartered class super user programs and resources for logical partitions running on any operating system protection Resources. In this class resources are affecting the overall resources of the main channel adapter. Just as management The trusted application can access a super class franchise resources.
While the preferred embodiment of the present invention described above is used with the four classes, but the present invention is not limited Thereto, can be made without departing from the spirit and scope of the present invention in case of use of any number of classes to use. For example, if the application does not have to support the virtual machine, you can cancel the actual address classes. Moreover, with the One or more of the above classes in combination with, or replaced with said one or more classes to use their He uses the class.
The use of the above-defined class provides can be assigned to the primary channel adapter access rights to resources Hierarchy. For example, super-privileged resources can only be the main channel adapter manager to access. Charter (And physical address) class resources can be allocated only by their operating system to partition and super pipe Manager access. User space resources can be their associated user program, the operating system partition and management Manager access. This hierarchical access structure is illustrated in the following Table 1, as described in Table 1 in order to access Ascending organization.
Use the classPermission to access
User● The user program ● Operating system partition ● Manager
Charter● Operating system partition ● Manager
Physical address● Operating system partition ● Manager This is independent of the class, is used to indicate translation of the relevant Included within the resources associated to the physical address. It is used to support virtual machine (VM) applications.
● Operating system partition ● Manager This is independent of the class, is used to indicate translation of the relevant Included within the resources associated to the physical address. It is used to support virtual machine (VM) applications....● Operating system partition ● Manager This is independent of the class, is used to indicate translation of the relevant Included within the resources associated to the physical address. It is used to support virtual machine (VM) applications....
Table 1 - using the class stratification
Has been established using the class stratification scheme, this scheme uses the class stratification through the queue right, Completion queue, event queue, and memory areas, and the system memory address space allocated to the memory pages Side to the primary channel adapter is implemented. That is, for each individual queue right, complete queues, event Queue facility and a memory area addresses are mapped to system memory address space different pages. For each queue pair, completion queue, event queue and memory area, a system memory address space The four pages are allocated - each using a class one. System memory address space is filled these pages Charge associated with them by using the class corresponding to the requester access to the queue pair, completion queue, Event queue or the facility address memory area. ...
Has been established using the class stratification scheme, this scheme uses the class stratification through the queue right, Completion queue, event queue, and memory areas, and the system memory address space allocated to the memory pages Side to the primary channel adapter is implemented. That is, for each individual queue right, complete queues, event Queue facility and a memory area addresses are mapped to system memory address space different pages. For each queue pair, completion queue, event queue and memory area, a system memory address space The four pages are allocated - each using a class one. System memory address space is filled these pages Charge associated with them by using the class corresponding to the requester access to the queue pair, completion queue, Event queue or the facility address memory area. ...
Figure 13 illustrates in accordance with the present invention, the system memory address space of the HCA in space BAR Examples. Shown in Figure 13, each queue pair, completion queue, event queue, and the memory region has The use of different classes for each specific system memory address space pages. BAR used for team Column, the completion queues, event queues, and user address space memory area, concession address space, Chartered real address space and the address space stored in ultra-associated base address of the page. Result, A preferred embodiment, there is a 16 BAR. According to BAR and access associated with BAR The BAR space offset to identify each specific page using the space. Each BAR space size BAR space is dependent on the particular queue pair, completion queue, event queue, or the number of memory areas Multiplied by the system memory address space page size. ...
Figure 14 illustrates in accordance with the present invention, the queue assigned to indicate the address space of the system memory page An expanded view of the surface. Shown in Figure 14, the system memory may be made for each page comprises an associated The use of the class of the queue to the requester access to the facility address lists. As shown, the user ground Address space using the class associated with the page provides a first set of queues to the facility address. Address with the concession Space associated page 2 provides a second set of queues to the facility address. Because the use of class is stratified into So as to have the address space license application and operating system than the user address space applications access to higher Q. Right, so the second set of queues for the facilities included in thepage 1 of the facility identified in the queue, plus only Address space can be used by a licensed class or higher application and operating system access on the other queue Facilities. ...
Figure 14 illustrates in accordance with the present invention, the queue assigned to indicate the address space of the system memory page An expanded view of the surface. Shown in Figure 14, the system memory may be made for each page comprises an associated The use of the class of the queue to the requester access to the facility address lists. As shown, the user ground Address space using the class associated with the page provides a first set of queues to the facility address. Address with the concession Space associated page 2 provides a second set of queues to the facility address. Because the use of class is stratified into So as to have the address space license application and operating system than the user address space applications access to higher Q. Right, so the second set of queues for the facilities included in thepage 1 of the facility identified in the queue, plus only Address space can be used by a licensed class or higher application and operating system access on the other queue Facilities. ...
Address space with the ultra-chartered pages associated four queues are provided in the fourth group of the facility address. The fourth group of the facility will include a queue for all queues on HCA facilities in the queue for the facility Address, because the address space is available super Chartered most privileged address space. Typically, such as manager The trusted code is just super license is assigned address space type of application used.
% E5% 9B% BE15% E5% 9B% BE% E8% A7% A3% E4% BA% 86% E5% BD% 93% E7% A1% AE% E5% AE% 9A% E5% AF% B9HCA% E8 % AE% BE% E6% 96% BD% E8% AE% BF% E9% 97% AE% E7% 9A% 84% E8% AF% B7% E6% B1% 82% E8% 80% 85% E6% 98 % AF% E5% 90% A6% E5% BA% 94% E5% BD% 93% E8% A2% AB% E6% 8F% 90% E4% BE% 9B% E5% AF% B9% E4% BA% 8EHCA % 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% E8% AE% BE% E6% 96% BD% E7% 9A% 84% E8% AE% BF % E9% 97% AE% E6% 97% B6% E6% 9C% AC% E5% 8F% 91% E6% 98% 8E% E7% 9A% 84% E6% 95% B4% E4% BD% 93% E6 % 93% 8D% E4% BD% 9C% E3% 80% 82% E5% A6% 82% E5% 9B% BE15% E6% 89% 80% E7% A4% BA% EF% BC% 8C% E5% BA % 94% E7% 94% A81510% E5% 90% 91% E4% B8% BB% E7% B3% BB% E7% BB% 9F% E5% A4% 84% E7% 90% 86% E5% 99% A8 % 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 201520% E5% 8F% 91% E9% 80% 81% E8% AE% BF% E9% 97% AE % E8% AF% B7% E6% B1% 82% E3% 80% 82% E8% AE% BF% E9% 97% AE% E8% AF% B7% E6% B1% 82% E5% 8C% 85% E6 % 8B% AC% E6% 9C% 9F% E6% 9C% 9B% E8% AE% BF% E9% 97% AE% E7% 9A% 84% E8% 99% 9A% E6% 8B% 9F% E5% 9C % B0% E5% 9D% 80% EF% BC% 8C% E8% AF% B8% E5% A6% 82% E5% AF% B9% E4% BA% 8E% E8% 99% 9A% E6% 8B% 9F % E5% 9C% B0% E5% 9D% 80% 0A% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% 20% E7% 9A% 84% E5% 86% 99 % E5% 85% A5% E6% 88% 96% E8% AF% BB% E5% 8F% 96% E6% 93% 8D% E4% BD% 9C% E3% 80% 82
The access request is the main system processor's virtual address translation process page table 1530. Virtually 1530 page table address translation performed in the page table lookup to determine the virtual address associated with the actual Address. 1510 then determines whether the application can access the actual address of the system memory address space Page. For example, this determination may be made either by determining the application identifier of a logical partition 1510 Corresponds to the physical address space of the system memory address associated with the page identifier of the logical partitions. Or Who can be known in the art using standard page access control mechanism to determine whether the application of 1610 The physical address can access the system memory address space pages. ...
If the application 1510 may access the system memory space of the physical address page, the actual address is Send to HCA 1540. HCA according to the requested type of access to inspect and each queue pair, complete Queues, event queue, or memory area associated with the page, wherein all of the requested type of access If the write queue for user facilities, the write queue for licensed facilities, write super franchise completion queue facilities. For example, HCA can be compared with the received BAR physical address. Because each of the different types of resources (QP, CQ, EQ, etc.) has its own set of four BAR, BAR matched by determining which The received address, HCA determine which resource is being accessed. ...
If the application 1510 may access the system memory space of the physical address page, the actual address is Send to HCA 1540. HCA according to the requested type of access to inspect and each queue pair, complete Queues, event queue, or memory area associated with the page, wherein all of the requested type of access If the write queue for user facilities, the write queue for licensed facilities, write super franchise completion queue facilities. For example, HCA can be compared with the received BAR physical address. Because each of the different types of resources (QP, CQ, EQ, etc.) has its own set of four BAR, BAR matched by determining which The received address, HCA determine which resource is being accessed. ...
If the application uses the class 1510 HCA sufficient access facilities, i.e. the use of the desired class or more High same, permission to access the HCA facilities. Otherwise, do not allow access to the HCA facilities.
If the application uses the class 1510 HCA sufficient access facilities, i.e. the use of the desired class or more High same, permission to access the HCA facilities. Otherwise, do not allow access to the HCA facilities....
If the application uses the class 1510 HCA sufficient access facilities, i.e. the use of the desired class or more High same, permission to access the HCA facilities. Otherwise, do not allow access to the HCA facilities....
By performing for the virtual address to a physical address translation page table lookup (step 1610) to Starting the process. According to the actual virtual address is mapped address, determine whether to allow access to the page (step Step 1620). If you allow access to the page, to the HCA to send the actual address (step 1630). For example, in the main system can perform the steps 1610-1630.
HCA requester according to the page address to identify the class (step 1640). Then, to determine whether the Access are sufficient for the facility to be addressed (step 1650). If the access are sufficient for the Addressing facilities, complete access (step 1660), and the process terminates thereafter.
In this returns to step 1650, if the access is not enough for the class to be addressed facilities, access is denied Q (step 1670), and the subsequent termination of treatment. In this see step 1620, if the page is not allowed Side access, the process proceeds to step 1670 as described above.
By using the present invention, a system in a logical partition prevents unauthorized access to the HCA Visit. The present invention, since the logical partition from one application and the operating system can not access the Different logical partitions HCA associated facilities, these applications and operating systems are prevented compromise and HCA facilities associated with these logical partitions applications and operating systems operation. Moreover, in the application, or The operation of the operating system failure or malfunction with the application or the operating system associated with a logical partition HCA isolated facilities, so other logical partitions are not affected.
It is important, though already fully functional data processing system within the scope of the present invention is described, However, those skilled in the art will understand that the process of the invention can be computer instruction R form and forms to distribute media, and the present invention can be equally applied irrespective of actually used The distribution of signal bearing media perform a specific type how. Examples of computer-readable medium, such as Floppy disks, hard drives, RAM, CD-ROM, DVD-ROM can record all types of media and If the input digital and analog communication links using transmission forms of wired or wireless communication link transmission class Type of media, such as radio frequency and form of the transmission light transmission. The computer-readable medium may be used Programme In the form of code format, the encoding format is decoded to the actual data for a particular processing system. ...
It is important, though already fully functional data processing system within the scope of the present invention is described, However, those skilled in the art will understand that the process of the invention can be computer instruction R form and forms to distribute media, and the present invention can be equally applied irrespective of actually used The distribution of signal bearing media perform a specific type how. Examples of computer-readable medium, such as Floppy disks, hard drives, RAM, CD-ROM, DVD-ROM can record all types of media and If the input digital and analog communication links using transmission forms of wired or wireless communication link transmission class Type of media, such as radio frequency and form of the transmission light transmission. The computer-readable medium may be used Programme In the form of code format, the encoding format is decoded to the actual data for a particular processing system. ...

Claims (36)

CNB2004100325476A2003-04-102004-04-08Equipment for controlling access of facilities according to the type of applicationExpired - Fee RelatedCN1310475C (en)

Applications Claiming Priority (2)

Application NumberPriority DateFiling DateTitle
US10/411,447US7010633B2 (en)2003-04-102003-04-10Apparatus, system and method for controlling access to facilities based on usage classes
US10/411,4472003-04-10

Publications (2)

Publication NumberPublication Date
CN1536842Atrue CN1536842A (en)2004-10-13
CN1310475C CN1310475C (en)2007-04-11

Family

ID=33130977

Family Applications (1)

Application NumberTitlePriority DateFiling Date
CNB2004100325476AExpired - Fee RelatedCN1310475C (en)2003-04-102004-04-08Equipment for controlling access of facilities according to the type of application

Country Status (2)

CountryLink
US (1)US7010633B2 (en)
CN (1)CN1310475C (en)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN101247393B (en)*2007-02-132011-10-12国际商业机器公司System and method for preventing IP spoofing and facilitating parsing of private data areas in system area network connection requests
CN102906702A (en)*2010-06-232013-01-30国际商业机器公司 Guest access to the adapter's address space
CN103562805A (en)*2011-03-302014-02-05维斯塔斯风力系统集团公司Wind power plant with highly reliable real-time power control
CN103582850A (en)*2011-03-302014-02-12维斯塔斯风力系统集团公司Distributed fault-tolerant control and protection system
CN103842976A (en)*2011-09-252014-06-04超威半导体公司Input/output memory management unit with protection mode for preventing memory access by i/o devices
CN104170348A (en)*2012-05-102014-11-26甲骨文国际公司System and method for supporting state synchronization in a network environment
CN104246742A (en)*2012-01-172014-12-24英特尔公司 Techniques for Command Verification for Remote Client Access to Storage Devices
US9134911B2 (en)2010-06-232015-09-15International Business Machines CorporationStore peripheral component interconnect (PCI) function controls instruction
US9195623B2 (en)2010-06-232015-11-24International Business Machines CorporationMultiple address spaces per adapter with address translation
US9213661B2 (en)2010-06-232015-12-15International Business Machines CorporationEnable/disable adapters of a computing environment
US9626298B2 (en)2010-06-232017-04-18International Business Machines CorporationTranslation of input/output addresses to memory addresses
US9634849B2 (en)2011-07-112017-04-25Oracle International CorporationSystem and method for using a packet process proxy to support a flooding mechanism in a middleware machine environment
US9935848B2 (en)2011-06-032018-04-03Oracle International CorporationSystem and method for supporting subnet manager (SM) level robust handling of unkown management key in an infiniband (IB) network
CN109641303A (en)*2016-08-312019-04-16依赛彼公司For realizing the technology of memory segmentation in welding or diced system

Families Citing this family (53)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
US7543290B2 (en)*2001-04-112009-06-02Mellanox Technologies Ltd.Multiple queue pair access with single doorbell
US7194540B2 (en)*2001-09-282007-03-20Intel CorporationMechanism for allowing multiple entities on the same host to handle messages of same service class in a cluster
US7111101B1 (en)*2003-05-072006-09-19Ayago Technologies General Ip (Singapore) Ptd. Ltd.Method and system for port numbering in an interconnect device
US7617376B2 (en)*2003-08-142009-11-10Hewlett-Packard Development Company, L.P.Method and apparatus for accessing a memory
US7979548B2 (en)*2003-09-302011-07-12International Business Machines CorporationHardware enforcement of logical partitioning of a channel adapter's resources in a system area network
US7552436B2 (en)*2003-11-252009-06-23International Business MachinesMemory mapped input/output virtualization
US7234037B2 (en)*2003-11-252007-06-19International Business Machines CorporationMemory mapped Input/Output operations
US7263568B2 (en)*2004-03-312007-08-28Intel CorporationInterrupt system using event data structures
US7197588B2 (en)*2004-03-312007-03-27Intel CorporationInterrupt scheme for an Input/Output device
US20060004983A1 (en)*2004-06-302006-01-05Tsao Gary YMethod, system, and program for managing memory options for devices
US20060095690A1 (en)*2004-10-292006-05-04International Business Machines CorporationSystem, method, and storage medium for shared key index space for memory regions
US7437447B2 (en)*2004-11-122008-10-14International Business Machines CorporationMethod and system for authenticating a requestor without providing a key
US8706942B2 (en)*2004-12-292014-04-22Intel CorporationDirect memory access (DMA) address translation between peer-to-peer input/output (I/O) devices
US7370174B2 (en)*2005-01-052008-05-06Intel CorporationMethod, system, and program for addressing pages of memory by an I/O device
US20060212870A1 (en)*2005-02-252006-09-21International Business Machines CorporationAssociation of memory access through protection attributes that are associated to an access control level on a PCI adapter that supports virtualization
US7657945B2 (en)*2005-03-022010-02-02International Business Machines CorporationSystems and arrangements to adjust resource accessibility based upon usage modes
US7853957B2 (en)*2005-04-152010-12-14Intel CorporationDoorbell mechanism using protection domains
US8635388B2 (en)*2006-03-312014-01-21Broadcom CorporationMethod and system for an OS virtualization-aware network interface card
US7996583B2 (en)*2006-08-312011-08-09Cisco Technology, Inc.Multiple context single logic virtual host channel adapter supporting multiple transport protocols
US7865633B2 (en)*2006-08-312011-01-04Cisco Technology, Inc.Multiple context single logic virtual host channel adapter
US7870306B2 (en)*2006-08-312011-01-11Cisco Technology, Inc.Shared memory message switch and cache
US7987469B2 (en)2006-12-142011-07-26Intel CorporationRDMA (remote direct memory access) data transfer in a virtual environment
US7624262B2 (en)*2006-12-202009-11-24International Business Machines CorporationApparatus, system, and method for booting using an external disk through a virtual SCSI connection
US7715428B2 (en)*2007-01-312010-05-11International Business Machines CorporationMulticore communication processing
US20080189432A1 (en)*2007-02-022008-08-07International Business Machines CorporationMethod and system for vm migration in an infiniband network
US7924829B2 (en)*2007-04-302011-04-12International Business Machines CorporationTechnique for enabling network statistics on software partitions
US8244826B2 (en)2007-10-232012-08-14International Business Machines CorporationProviding a memory region or memory window access notification on a system area network
US8331362B2 (en)*2008-12-302012-12-11Juniper Networks, Inc.Methods and apparatus for distributed dynamic network provisioning
US8565118B2 (en)2008-12-302013-10-22Juniper Networks, Inc.Methods and apparatus for distributed dynamic network provisioning
US8255496B2 (en)2008-12-302012-08-28Juniper Networks, Inc.Method and apparatus for determining a network topology during network provisioning
US8054832B1 (en)2008-12-302011-11-08Juniper Networks, Inc.Methods and apparatus for routing between virtual resources based on a routing location policy
US8190769B1 (en)2008-12-302012-05-29Juniper Networks, Inc.Methods and apparatus for provisioning at a network device in response to a virtual resource migration notification
US8953603B2 (en)2009-10-282015-02-10Juniper Networks, Inc.Methods and apparatus related to a distributed switch fabric
US8442048B2 (en)2009-11-042013-05-14Juniper Networks, Inc.Methods and apparatus for configuring a virtual network switch
US9507619B2 (en)*2010-01-122016-11-29International Business Machines CorporationVirtualizing a host USB adapter
US8891406B1 (en)2010-12-222014-11-18Juniper Networks, Inc.Methods and apparatus for tunnel management within a data center
WO2013028175A1 (en)*2011-08-232013-02-28Intel CorporationGid capable switching in an infiniband fabric
US9208319B2 (en)2011-12-152015-12-08Microsoft Technology Licensing, LlcCode base partitioning system
US8949199B2 (en)*2011-12-292015-02-03Dell Products L.P.Systems and methods for de-duplication in storage systems
US9231888B2 (en)2012-05-112016-01-05Oracle International CorporationSystem and method for routing traffic between distinct InfiniBand subnets based on source routing
US9665719B2 (en)2012-06-042017-05-30Oracle International CorporationSystem and method for supporting host-based firmware upgrade of input/output (I/O) devices in a middleware machine environment
US9503093B2 (en)*2014-04-242016-11-22Xilinx, Inc.Virtualization of programmable integrated circuits
WO2016118128A1 (en)*2015-01-222016-07-28Hewlett Packard Enterprise Development LpRouter to send a request from a first subnet to a second subnet
US10013360B2 (en)2015-03-042018-07-03Cavium, Inc.Managing reuse information with multiple translation stages
US9720773B2 (en)*2015-03-042017-08-01Cavium, Inc.Managing reuse information in caches
US9471509B2 (en)*2015-03-192016-10-18Cavium, Inc.Managing address-independent page attributes
US11018947B2 (en)2016-01-272021-05-25Oracle International CorporationSystem and method for supporting on-demand setup of local host channel adapter port partition membership in a high-performance computing environment
US10972375B2 (en)2016-01-272021-04-06Oracle International CorporationSystem and method of reserving a specific queue pair number for proprietary management traffic in a high-performance computing environment
US10594547B2 (en)2016-01-272020-03-17Oracle International CorporationSystem and method for application of virtual host channel adapter configuration policies in a high-performance computing environment
US10296264B2 (en)2016-02-092019-05-21Samsung Electronics Co., Ltd.Automatic I/O stream selection for storage devices
US10873564B2 (en)*2018-09-202020-12-22Palo Alto Research Center IncorporatedCloud-based device manager based on message queues
US11874767B2 (en)2021-12-152024-01-16Hewlett Packard Enterprise Development LpMemory partitions for processing entities
US11768772B2 (en)2021-12-152023-09-26Hewlett Packard Enterprise Development LpAccumulators corresponding to bins in memory

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
DK0443983T3 (en)*1990-02-191996-03-18Ciba Geigy Ag acrylic Compounds
US6718392B1 (en)*2000-10-242004-04-06Hewlett-Packard Development Company, L.P.Queue pair partitioning in distributed computer system
US6917987B2 (en)*2001-03-262005-07-12Intel CorporationMethodology and mechanism for remote key validation for NGIO/InfiniBand™ applications
US7363389B2 (en)*2001-03-292008-04-22Intel CorporationApparatus and method for enhanced channel adapter performance through implementation of a completion queue engine and address translation engine
US6480500B1 (en)*2001-06-182002-11-12Advanced Micro Devices, Inc.Arrangement for creating multiple virtual queue pairs from a compressed queue pair based on shared attributes
US6834332B2 (en)*2001-08-302004-12-21International Business Machines CorporationApparatus and method for swapping-out real memory by inhibiting i/o operations to a memory region and setting a quiescent indicator, responsive to determining the current number of outstanding operations
US6804729B2 (en)*2002-09-302004-10-12International Business Machines CorporationMigrating a memory page by modifying a page migration state of a state machine associated with a DMA mapper based on a state notification from an operating system kernel
US20040193833A1 (en)*2003-03-272004-09-30Kathryn HamptonPhysical mode addressing

Cited By (28)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN101247393B (en)*2007-02-132011-10-12国际商业机器公司System and method for preventing IP spoofing and facilitating parsing of private data areas in system area network connection requests
US9342352B2 (en)2010-06-232016-05-17International Business Machines CorporationGuest access to address spaces of adapter
US9213661B2 (en)2010-06-232015-12-15International Business Machines CorporationEnable/disable adapters of a computing environment
CN102906702A (en)*2010-06-232013-01-30国际商业机器公司 Guest access to the adapter's address space
US9383931B2 (en)2010-06-232016-07-05International Business Machines CorporationControlling the selectively setting of operational parameters for an adapter
CN102906702B (en)*2010-06-232015-11-25国际商业机器公司 Guest access to the adapter's address space
US9626298B2 (en)2010-06-232017-04-18International Business Machines CorporationTranslation of input/output addresses to memory addresses
US9134911B2 (en)2010-06-232015-09-15International Business Machines CorporationStore peripheral component interconnect (PCI) function controls instruction
US9195623B2 (en)2010-06-232015-11-24International Business Machines CorporationMultiple address spaces per adapter with address translation
CN103562805A (en)*2011-03-302014-02-05维斯塔斯风力系统集团公司Wind power plant with highly reliable real-time power control
US9983555B2 (en)2011-03-302018-05-29Vestas Wind Systems A/SDistributed fault-tolerant control and protection system
CN103582850A (en)*2011-03-302014-02-12维斯塔斯风力系统集团公司Distributed fault-tolerant control and protection system
US9935848B2 (en)2011-06-032018-04-03Oracle International CorporationSystem and method for supporting subnet manager (SM) level robust handling of unkown management key in an infiniband (IB) network
US9641350B2 (en)2011-07-112017-05-02Oracle International CorporationSystem and method for supporting a scalable flooding mechanism in a middleware machine environment
US9634849B2 (en)2011-07-112017-04-25Oracle International CorporationSystem and method for using a packet process proxy to support a flooding mechanism in a middleware machine environment
CN103842976B (en)*2011-09-252016-01-20超威半导体公司There is protected mode with the I/O Memory Management Unit preventing I/O device from carrying out memory access
CN103842976A (en)*2011-09-252014-06-04超威半导体公司Input/output memory management unit with protection mode for preventing memory access by i/o devices
CN104246742B (en)*2012-01-172017-11-10英特尔公司 Techniques for Command Verification for Remote Client Access to Storage Devices
CN104246742A (en)*2012-01-172014-12-24英特尔公司 Techniques for Command Verification for Remote Client Access to Storage Devices
US9529878B2 (en)2012-05-102016-12-27Oracle International CorporationSystem and method for supporting subnet manager (SM) master negotiation in a network environment
US9690836B2 (en)2012-05-102017-06-27Oracle International CorporationSystem and method for supporting state synchronization in a network environment
US9690835B2 (en)2012-05-102017-06-27Oracle International CorporationSystem and method for providing a transactional command line interface (CLI) in a network environment
US9594818B2 (en)2012-05-102017-03-14Oracle International CorporationSystem and method for supporting dry-run mode in a network environment
US9852199B2 (en)2012-05-102017-12-26Oracle International CorporationSystem and method for supporting persistent secure management key (M—Key) in a network environment
CN104170348B (en)*2012-05-102018-02-13甲骨文国际公司The system and method synchronous for status of support in a network environment
US9563682B2 (en)2012-05-102017-02-07Oracle International CorporationSystem and method for supporting configuration daemon (CD) in a network environment
CN104170348A (en)*2012-05-102014-11-26甲骨文国际公司System and method for supporting state synchronization in a network environment
CN109641303A (en)*2016-08-312019-04-16依赛彼公司For realizing the technology of memory segmentation in welding or diced system

Also Published As

Publication numberPublication date
US7010633B2 (en)2006-03-07
US20040205253A1 (en)2004-10-14
CN1310475C (en)2007-04-11

Similar Documents

PublicationPublication DateTitle
CN1536842A (en)Equipment for controlling access of facilities according to the type of application
JP3783017B2 (en) End node classification using local identifiers
CN1604057A (en)Method and system for hardware enforcement of logical partitioning of a channel adapter's resources in a system area network
CN111221758B (en) Method and computer device for processing remote direct memory access requests
US10924483B2 (en)Packet validation in virtual network interface architecture
US7493409B2 (en)Apparatus, system and method for implementing a generalized queue pair in a system area network
US7543290B2 (en)Multiple queue pair access with single doorbell
US6938138B2 (en)Method and apparatus for managing access to memory
CN1280725C (en) Computing systems and methods for resource balancing in a partitioned processing environment
US6883099B2 (en)Secure virtual interface
US6718392B1 (en)Queue pair partitioning in distributed computer system
CN114025021B (en)Communication method, system, medium and electronic equipment crossing Kubernetes cluster
US8370447B2 (en)Providing a memory region or memory window access notification on a system area network
US7996569B2 (en)Method and system for zero copy in a virtualized network environment
CN1617526A (en)Method and device for emulating multiple logic port on a physical poet
EP2725518A1 (en)Method and apparatus for securing and segregating host to host messaging on pcie fabric
US20040202189A1 (en)Apparatus, system and method for providing multiple logical channel adapters within a single physical channel adapter in a systen area network
US20150096051A1 (en)Method, Apparatus, and System for Manageability and Secure Routing and Endpoint Access
CN1640089A (en)Methodology and mechanism for remote key validation for NGIO/InfiniBand applications
US10872056B2 (en)Remote memory access using memory mapped addressing among multiple compute nodes
CN1487418A (en)Far-end divect memory access invocating memory management unloading of network adapter
US11036649B2 (en)Network interface card resource partitioning
JP2004531001A (en) Data transfer between host computer system and Ethernet adapter
US8316156B2 (en)Method and apparatus for interfacing device drivers to single multi-function adapter
US8291176B2 (en)Protection domain groups to isolate access to memory windows

Legal Events

DateCodeTitleDescription
C06Publication
PB01Publication
C10Entry into substantive examination
SE01Entry into force of request for substantive examination
C14Grant of patent or utility model
GR01Patent grant
CF01Termination of patent right due to non-payment of annual fee

Granted publication date:20070411


[8]ページ先頭

©2009-2025 Movatter.jp